1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
|
/***************************************************************
* _ *
* | | *
* __ _ ___ _ __ ___ _ __ ___ ___ | | ___ __ _ _ _ *
* / _` |/ _ \ '_ ` _ \| '_ ` _ \ / _ \| |/ _ \ / _` | | | | *
* | (_| | __/ | | | | | | | | | | (_) | | (_) | (_| | |_| | *
* \__, |\___|_| |_| |_|_| |_| |_|\___/|_|\___/ \__, |\__, | *
* __/ | __/ | __/ | *
* |___/ |___/ |___/ *
* *
* version 0.1 *
***************************************************************/
#ifndef GEMMOLOGY_FWD_H
#define GEMMOLOGY_FWD_H
#include <cstdint>
#include <cstring>
#include <tuple>
#include <xsimd/xsimd.hpp>
namespace gemmology {
namespace callbacks {
struct Unquantize {
float unquant_mult;
template <class Arch>
xsimd::batch<float, Arch> operator()(xsimd::batch<int32_t, Arch> total, size_t, size_t, size_t);
template <class Arch>
std::tuple<xsimd::batch<float, Arch>, xsimd::batch<float, Arch>> operator()(
std::tuple<xsimd::batch<int32_t, Arch>, xsimd::batch<int32_t, Arch>>
total,
size_t, size_t, size_t);
};
struct AddBias {
const float *bias_addr;
template <class Arch>
xsimd::batch<float, Arch> operator()(xsimd::batch<float, Arch> total, size_t, size_t col_idx,
size_t);
template <class Arch>
std::tuple<xsimd::batch<float, Arch>, xsimd::batch<float, Arch>>
operator()(
std::tuple<xsimd::batch<float, Arch>, xsimd::batch<float, Arch>> total,
size_t, size_t col_idx, size_t);
};
struct Write {
float *output_addr;
Write(float *o) : output_addr(o) {}
template <class Arch>
void operator()(xsimd::batch<float, Arch> result, size_t row_idx,
size_t col_idx, size_t col_size);
template <class Arch>
void operator()(xsimd::batch<int32_t, Arch> result, size_t row_idx,
size_t col_idx, size_t col_size);
template <class Arch>
void operator()(
std::tuple<xsimd::batch<float, Arch>, xsimd::batch<float, Arch>> result,
size_t row_idx, size_t col_idx, size_t col_size);
template <class Arch>
void operator()(
std::tuple<xsimd::batch<int32_t, Arch>, xsimd::batch<int32_t, Arch>>
result,
size_t row_idx, size_t col_idx, size_t col_size);
};
struct UnquantizeAndWrite {
Unquantize unquantize;
Write write;
UnquantizeAndWrite(float factor, float *output)
: unquantize{factor}, write{output} {}
template <class T>
void operator()(T const &total, size_t row_idx, size_t col_idx,
size_t col_size);
};
struct UnquantizeAndAddBiasAndWrite {
Unquantize unquantize;
AddBias add_bias;
Write write;
UnquantizeAndAddBiasAndWrite(float factor, const float *bias, float *output)
: unquantize{factor}, add_bias{bias}, write{output} {}
template <class T>
void operator()(T const &total, size_t row_idx, size_t col_idx,
size_t col_size);
};
} // namespace callbacks
//
// Arch-specific implementation of each routine
//
template <class Arch> struct Engine {
static void QuantizeU(const float *input, uint8_t *output, float quant_mult,
size_t size);
static void Quantize(const float *const input, int8_t *const output,
float quant_mult, size_t size);
template <typename IntegerTy>
static void SelectColumnsB(const int8_t *input, int8_t *output, size_t rows,
const IntegerTy *cols_begin,
const IntegerTy *cols_end);
static void PrepareBTransposed(const float *input, int8_t *output,
float quant_mult, size_t cols, size_t rows);
static void PrepareBQuantizedTransposed(const int8_t *input, int8_t *output,
size_t cols, size_t rows);
static void PrepareB(const float *input, int8_t *output_shadow,
float quant_mult, size_t rows, size_t cols);
static void PrepareA(const float *input, int8_t *output, float quant_mult,
size_t rows, size_t cols);
struct Shift {
static void PrepareA(const float *input, uint8_t *output, float quant_mult,
size_t rows, size_t cols);
template <class Callback>
static void Multiply(const uint8_t *A, const int8_t *B, size_t A_rows,
size_t width, size_t B_cols, Callback callback);
template <class Callback>
static void PrepareBias(const int8_t *B, size_t width, size_t B_cols,
Callback C);
};
};
//
// Top-level wrappers that mostly match intgemm API
//
template <class Arch = xsimd::default_arch>
inline void QuantizeU(const float *input, uint8_t *output, float quant_mult,
size_t size) {
return Engine<Arch>::QuantizeU(input, output, quant_mult, size);
}
template <class Arch = xsimd::default_arch>
inline void Quantize(const float *const input, int8_t *const output,
float quant_mult, size_t size) {
return Engine<Arch>::Quantize(input, output, quant_mult, size);
}
template <class Arch = xsimd::default_arch, typename IntegerTy>
inline void SelectColumnsB(const int8_t *input, int8_t *output, size_t rows,
const IntegerTy *cols_begin,
const IntegerTy *cols_end) {
return Engine<Arch>::SelectColumnsB(input, output, rows, cols_begin,
cols_end);
}
template <class Arch = xsimd::default_arch>
inline void PrepareBTransposed(const float *input, int8_t *output,
float quant_mult, size_t cols, size_t rows) {
return Engine<Arch>::PrepareBTransposed(input, output, quant_mult, cols,
rows);
}
template <class Arch = xsimd::default_arch>
inline void PrepareBQuantizedTransposed(const int8_t *input, int8_t *output,
size_t cols, size_t rows) {
return Engine<Arch>::PrepareBQuantizedTransposed(input, output, cols, rows);
}
template <class Arch = xsimd::default_arch>
inline void PrepareB(const float *input, int8_t *output_shadow,
float quant_mult, size_t rows, size_t cols) {
return Engine<Arch>::PrepareB(input, output_shadow, quant_mult, rows, cols);
}
template <class Arch = xsimd::default_arch>
inline void PrepareA(const float *input, int8_t *output, float quant_mult,
size_t rows, size_t cols) {
return Engine<Arch>::PrepareA(input, output, quant_mult, rows, cols);
}
namespace Shift {
template <class Arch = xsimd::default_arch>
inline void PrepareA(const float *input, uint8_t *output, float quant_mult,
size_t rows, size_t cols) {
return Engine<Arch>::Shift::PrepareA(input, output, quant_mult, rows, cols);
}
template <class Arch = xsimd::default_arch, class Callback>
inline void Multiply(const uint8_t *A, const int8_t *B, size_t A_rows,
size_t width, size_t B_cols, Callback C) {
return Engine<Arch>::Shift::Multiply(A, B, A_rows, width, B_cols, C);
}
template <class Arch = xsimd::default_arch, class Callback>
inline void PrepareBias(const int8_t *B, size_t width, size_t B_cols,
Callback C) {
return Engine<Arch>::Shift::PrepareBias(B, width, B_cols, C);
}
} // namespace Shift
} // namespace gemmology
#endif
|
